Do IPO Firms Purchase Analyst Coverage With Underpricing?*
Michael T. Cliff and David J. Denis
Krannert Graduate School of Management Purdue University
West Lafayette, IN 47907-1310
[email protected] [email protected]
September, 2003
1
Do IPO Firms Purchase Analyst Coverage With Underpricing?
Abstract
We examine the links among IPO underpricing, post-IPO analyst coverage, and the likelihood of
switching underwriters. Our findings indicate that underpricing is positively related to analyst
coverage by the lead underwriter and to the presence of an all-star analyst on the research staff of
the lead underwriter. These findings are robust to controls for other determinants of
underpricing previously documented in the literature and to controls for the endogeneity of
underpricing and analyst coverage. In addition, after controlling for other potential determinants
of switching underwriters, we find that the probability of switching underwriters between IPO
and SEO is negatively related to the unexpected amount of post-IPO analyst coverage. We
interpret these findings as consistent with the hypothesis that underpricing is, in part,
compensation for expected post-IPO analyst coverage from highly ranked analysts.
2
Do IPO Firms Purchase Analyst Coverage With Underpricing?
Investment bankers provide a wide range of services to firms issuing new shares through
an initial public offering (IPO). These services include pre-IPO activities related to the pricing,
marketing, and distribution of the offering, as well as post-IPO activities such as price
stabilization, market making, and analyst research coverage. Despite the variety of services
provided to issuers and the variation in issuer characteristics, there is surprisingly little variation
in the direct costs of completing an IPO. Chen and Ritter (2000) and Hansen (2001) show that
underwriter spreads in IPOs are clustered at 7% for all but the very smallest and very largest
offerings. Moreover, a 15% overallotment option is a standard feature of IPO contracts.
Both anecdotal and academic evidence indicates that research coverage has become an
essential element of the security issuance process in recent years. Press reports indicate that star
analysts play an important role in securing underwriting business.1 This view is confirmed by
Dunbar (2000), who reports that underwriters increase their market share of IPOs if they have an
analyst highly rated in the annual Institutional Investor survey, and Clarke, Dunbar, and Kahle
(2003), who report that underwriters adding an all-star analyst gain greater IPO market share
(though losing an all-star is not associated with a decline in market share). Further confirmation
of the importance of research coverage in the choice of underwriter is provided by Krigman,
Shaw, and Womack (2001). Krigman et al. report survey evidence indicating that improved
research coverage is the most important element of the decision to switch underwriters between a
company’s IPO and its subsequent seasoned equity offering (SEO). The bottom line is that
isssuing companies appear to place a value on securing research coverage from sell-side analysts,
especially those who are highly-ranked.2
If companies value research coverage, it follows that they are willing to allocate resources
to acquire this coverage. Yet it is unclear how the payment for such service is made in IPOs. In
this study, we empirically examine the hypothesis that issuing firms pay for analyst coverage via
the underpricing of the offering. Lead underwriters can benefit from underpricing by allocating
3
IPOs to preferred clients (perhaps in exchange for future investment banking business or high
future trading commissions) and by serving as the primary market maker for the high aftermarket
trading volume that typically follows underpriced IPOs. Thus, we hypothesize that issuers
purchase analyst coverage by giving up greater underpricing at the time of the IPO. A corollary
of this hypothesis is that if the lead underwriter does not deliver the expected research coverage,
the issuing company is more likely to switch to a new underwriter for subsequent seasoned
equity offers (SEOs). Although ours is not the first study to examine the relation between
analyst research coverage and IPO underpricing, nor the first to examine the link between analyst
coverage and the decision to switch underwriters, we are, to our knowledge, the first to examine
the interconnections among these three aspects of the equity issuance process.
Our sample consists of 1,050 firms completing initial public offerings (IPOs) between
1993 and 2000 and also completing at least one subsequent SEO. We find that the analysts of
lead underwriters make post-IPO recommendations in 839 of the 1,050 offerings. Of these 839
recommendations, 793 (95%) are either strong buy or buy recommendations. Despite the
apparent uniformity in buy recommendations, however, there is a strong correlation between IPO
underpricing and both the frequency and the perceived quality of subsequent recommendations.
For companies in the lowest quintile of IPO underpricing, the lead underwriter makes a
recommendation (possibly including unfavorable ones) only 75% of the time. This rate increases
to 86% for the highest quintile of underpricing. The difference is significant at the 0.01 level.
Similarly, the lead underwriter has an all-star analyst (as defined by Institutional Investor)
following the industry of the IPO firm in 16% of the firms in the lowest quintile of underpricing.
This rises to 35% for the firms in the highest quintile of underpricing. These findings from
univariate tests are robust to controls for other determinants of underpricing and continue to hold
when we control for endogeneity using a two-stage procedure.
The positive relation between underpricing and analyst coverage is consistent with the
hypothesis that issuing firms compensate investment banks for high-quality analyst coverage via
the underpricing of the offering. That is, issuers knowingly choose an underwriter with a highly
4
ranked analyst with the expectation that there will be more money left on the table than if they
had chosen a different underwriter. This is consistent with Loughran and Ritter’s (2002b)
analyst lust hypothesis. An alternative (though not mutually exclusive) explanation, offered by
Aggarwal, Krigman, and Womack (2002), is that managers strategically underprice IPOs in
order to attract interest from analysts and the media, thereby building price momentum.
Our analysis of the likelihood that an IPO issuer will switch lead underwriters between its
IPO and its SEO helps distinguish the analyst lust hypothesis from the strategic underpricing
hypothesis. Although we confirm Krigman, Shaw and Womack’s (2001) finding that firms with
lower underpricing are more likely to switch underwriters, we find that, controlling for
underpricing, issuing companies are significantly more likely to switch lead underwriters if the
lead underwriter does not have a recommendation outstanding at the one-year anniversary of the
IPO. To our knowledge, the strategic underpricing hypothesis makes no predictions regarding
the relation between analyst coverage and the likelihood of switching underwriters. Collectively,
therefore, we believe our findings are most consistent with the hypothesis that underpricing is, in
part, compensation for expected post-IPO analyst coverage. If underwriters do not deliver the
expected analyst coverage (conditional on underpricing), the IPO firm is more likely to switch
underwriters when it issues shares in its subsequent SEO.
The remainder of the paper is organized as follows. In section I, we detail our testable
hypotheses and discuss how our study relates to other recent studies that examine IPO
underpricing and post-IPO analyst coverage. Section II describes our sample and experimental
design. Section III describes our main emprical results. Section IV discusses the implications of
our findings and offers concluding remarks.
I. Hypothesis Development and Relation to Prior Studies
We hypothesize that issuing companies purchase analyst coverage by deliberately
underpricing the IPO. In this section, we develop this and other hypotheses and discuss how our
study relates to prior work in the IPO literature.
5
A. Hypotheses
A necessary condition for the hypothesized link between underpricing and analyst
coverage is that analyst recommendations are perceived by issuing companies to be valuable.
Analyst recommendations might be valuable for several reasons. First, analyst coverage can
generate publicity for the issuing company, thereby potentially increasing firm value by
generating more customers.3 Second, both Chen and Ritter (2000) and Aggarwal, Krigman, and
Womack (2002) note that post-IPO analyst recommendations that boost share price can be
especially important for insiders wishing to sell their shares in the open market following
expiration of the lock-up period.4 Third, greater analyst coverage might lead to greater investor
recognition of the IPO company. According to Merton’s (1987) model, this greater investor
recognition can lead to a higher company value.
Loughran and Ritter (2002b) argue that analyst coverage has become more important to
issuers over time. They base this argument on three observations: (i) The use of co-managers in
IPO underwriting has increased over time. According to Loughran and Ritter, investment
bankers claim that co-managers are present in underwriting syndicates almost exclusively to
provide additional research coverage; (ii) Growth options have become a larger percentage of
firm value, thereby increasing the importance of analyst’s forecasts of future growth, and (iii)
Analysts are increasingly more visible via the internet and cable television.
Analyst recommendations are costly to the underwriter to provide. These costs include
not only the direct costs of investigation, but also any reputation costs associated with incorrect
recommendations. This implies that underwriters will, ceteris paribus, demand greater
compensation to underwrite deals that are subsequently accompanied by greater, more reputable,
or more favorable analyst coverage. One way to compensate underwriters for greater analyst
coverage would be to increase the underwriter fee. However, the fact that underwriter fees are a
uniform 7% for the majority of IPOs during our sample period (75% of our sample) suggests that
differential underwriter fees are not used as compensation for differential analyst coverage. We
6
hypothesize, therefore, that underwriters are compensated for analyst coverage via greater
underpricing.
Why wouldn’t firms compensate underwriters for analyst coverage via the underwriter
spread? One possibility is that uniform underwriter fees offer unique economic advantages in
serving IPOs. Hansen (2001) offers several conjectures as to why the 7% underwriter fee has
evolved as an efficient contract. These include reduced information externalities that arise is
valuing IPOs, reduced moral hazard in underwriter placement efforts, and lower contracting
costs. Alternatively, for reasons described below, underwriters may perceive greater benefits
from receiving compensation in the form of underpricing.
There are several ways in which underwriters might benefit from underpricing. First,
underwriters can allocate more underpriced IPOs to favored clients, perhaps in return for future
investment banking business. According to this hypothesis, labeled the corruption hypothesis by
Loughran and Ritter (2002b), the money left on the table in an underpriced deal is currency with
which investment bankers can compensate other venture capitalists and issuing company
executives. This practice, known as spinning, has been the subject of recent congressional
investigations of CSFB, Goldman Sachs, and Salomon-Smith Barney. The recently proposed
NASD Rule 2712 clarifies and strengthens the prior Rule 2710 which prohibits spinning.5
Second, underwriters can allocate shares to hedge funds and other large investors who then do
more of their trading with the investment bank. Some claim that these investors pay higher than
normal commissions.6 Third, because underpricing is positively correlated with subsequent
trading volume [Krigman, Shaw, and Womack (2001)] and lead underwriters are the primary
market makers [Ellis, Michaely and O’Hara (2000)], underwriting firms can benefit from
underpricing.
This discussion leads to several empirical predictions. First, we hypothesize that analyst
coverage by the lead underwriter is positively related to initial underpricing. While coverage can
be measured in several ways, our analysis focuses on (i) the existence of analyst
recommendations by lead underwriters, and (ii) the perceived quality of the lead underwriter’s
7
analyst. We focus on lead underwriters because they have the most to gain from underpricing
through their primary role in allocating IPOs and through their subsequent role as the primary
market makers. We focus on analyst recommendations rather than short-term earnings forecasts
because recommendations are longer term and, hence, more difficult to compare to actual
outcomes. Presumably, reputation effects will constrain analyst forecasts of near-term earnings
to be close to actual outcomes. Consistent with this conjecture, Lin and McNichols (1998) report
significant differences in the recommendations of lead underwriters of seasoned equity offerings
versus those of unaffiliated analysts, but no evidence of differences in short-term earnings
forecasts.
Second, we hypothesize that underwriters from investment banks with higher research
reputations demand greater underpricing as compensation for their services (i.e. they earn rents).
That is, conditional on making a recommendation, underpricing should be greater in IPOs
underwritten by more prestigious investment banks or those with higher rated analysts.
Third, we hypothesize that the likelihood of switching underwriters between the
company’s IPO and its SEO is associated with the unexpected amount of analyst coverage. That
is, if analysts do not deliver the expected coverage (conditional on underpricing), companies are
more likely to switch to a different underwriter for their SEO.
B. Relation to Prior Studies
At least three prior studies report a positive correlation between underpricing and some
measure of analyst coverage. Rajan and Servaes (1997) find that, controlling for the post-IPO
market value of equity, the number of analysts following an IPO stock is positively related to
underpricing. This finding is consistent with Chemmanur (1993), who predicts that equilibrium
offer prices may involve underpricing in order to maximize outsider information production. In
other words, unlike our hypothesis, Chemmanur’s (1993) model predicts that the direction of
causality runs from underpricing to analyst coverage. Similarly, Bradley, Jordan, and Ritter
(2003) find that the likelihood of coverage being initiated following the expiration of the so-
8
called “quiet period” is positively related to the degree of underpricing. However, their focus is
on the stock price reaction to the analyst recommendations.
Aggarwal, Krigman, and Womack (2002) find that underpricing is positively correlated
with analyst research coverage by non-lead underwriters. However, their focus is on testing the
hypothesis that managers strategically underprice to maximize the proceeds from open market
sales following the expiration of the lockup period. In other words, their study emphasizes the
benefits to issuing company managers from underpricing. In contrast, our study focuses on
analyst coverage of the lead underwriter and emphasizes potential benefits to the underwriter
from underpricing.
Other studies establish that post-IPO analyst coverage is typically abnormally favorable,
particularly for lead underwriters. For example, Bradley, Jordan, and Ritter (2003) report that
when analyst coverage is initiated, it is almost always with a favorable recommendation.
Michaely and Womack (1999) study a sample of 391 IPOs from 1990-1991 and report that lead
underwriters are significantly more likely to issue buy recommendations in the year following
the IPO than are non-lead underwriters. However, long-run performance following lead bank
recommendations is inferior to that following the recommendations of other banks. These
studies do not, however, investigate the link between underpricing and analyst coverage, nor do
they test whether this link affects the likelihood of switching underwriters in the company’s
subsequent SEO.
Krigman, Shaw, and Womack (2001) investigate the reasons why firms switch
underwriters for their SEO. Based on large-sample and survey evidence, they conclude that the
timeliness and perceived quality of research coverage is an important determinant of the decision
to switch. However, they do not investigate underpricing as a means of compensation for this
research coverage. In fact, they conclude that issuing companies “allocate their resources in the
form of underwriting fees, to increase and improve this coverage.” Because underwriting fees do
not vary much across issues, it is not clear how fees are used as compensation for differential
research coverage.
9
II. Sample Selection and Data Description
A. Sample Formation
We obtain our sample of issuing firms by first selecting all firms on the Securities Data
Corporation (SDC) New Issues database that completed an initial public offering between 1993
and 2000. Because we are interested in the dynamics of the relations among underpricing,
analyst recommendations, and subsequent underwriter choice, we also require that the sample
firms complete at least one seasoned equity offering. We then match these firms against the
Center for Research and Securities Prices (CRSP) and I/B/E/S databases. We exclude financial
firms (SIC codes 6xxx), firms that SDC lists as having multiple IPOs or concurrent offers, and
issues with SDC share types other than {‘Common Shares’, ‘Class A Shares’, ‘Ordinary Shares’,
or ‘Ord./Common Shrs’}. We also exclude nine offers for which Merrill Lynch is the lead
underwriter in 1993 and 1994.7 This results in a final sample of 1,050 IPOs during this period.
Although we choose the sample period of 1993 to 2000 to maximize the availability of
analyst recommendations on I/B/E/S, Bradley, Jordan, and Ritter (2003) report that I/B/E/S
coverage is less complete in the early years of our sample period. This raises the possibility that
we label some firms as having received no analyst coverage when, in fact, they did receive
coverage. Although we are unaware of any reason why such errors would be systematically
related to underpricing, we later test the robustness of our findings to the exclusion of offerings
completed in the first part of our sample period – i.e. the years in which the likelihood of errors
in the recording of analyst coverage is greatest.
By imposing the requirement that the sample firms complete at least one seasoned equity
offering, we potentially bias the sample towards more successful companies. If analysts are
more likely to cover successful companies, this increases the likelihood that our sample
companies will receive analyst coverage. Note, however, that, if anything, this lack of dispersion
in analyst coverage will make it less likely that we find any connection between IPO
underpricing and analyst coverage. Moreover, as we later show in Table I, the sample IPOs
10
exhibit levels of underpricing that are quite similar to that of the population of IPOs issued
during the same time period.
We use CRSP for data on share prices, including the initial trading price, and trading
volume. From SDC, we identify the lead underwriter(s) for each offering and attempt to find
I/B/E/S coverage of the issuer by that investment bank. In all of our analysis we make an effort
to match investment banks taking into account acquisitions. For example, Bankers Trust
acquired Alex. Brown in 1997. For an IPO in 1996 with Alex. Brown as the lead underwriter,
we would consider analyst coverage by both Alex. Brown and Bankers Trust in 1997. For an
IPO done by Alex. Brown in 1995, we would not consider Bankers Trust as affiliated with the
lead underwriter in 1996. We are able to determine a match for 96% of the issues in our sample.
Those IPOs for which we are not able to find a match are treated as if there is no analyst
coverage. For IPOs that have joint lead managers (i.e. more than one underwriter that help
manage the book - SDC codes ‘BM’, ‘JB’, or ‘LM’) we treat all lead managers as one. We do
not treat co-managers as the lead, however, since these underwriters are not book-runners,
leaving the lead manager to allocate the vast majority of shares [see Chen and Ritter (2000,
Table V)].
B. Variable Construction
The Appendix provides a summary of the key variables used in our analysis and the data
sources. We briefly discuss some of the most important variables here. We measure
underpricing as the percentage return from the SDC offer price to the first closing price on
CRSP. If the first CRSP price is more than three days after the SDC issue date, we delete the
issuer.
Measuring analyst coverage requires some subjective decisions on our part. Ideally, our
measure will indicate whether the lead underwriter provides research coverage that is both timely
and ongoing. Our primary measure is a dummy variable indicating whether the lead underwriter
provides a recommendation on the issuer one year after the IPO.8 Throughout the paper, when
11
we refer to a company receiving “coverage,” we are referring to this measure. We also consider
the strength of the recommendation, but since 95% of the leads’ recommendations are Strong
Buy or Buy, we focus primarily on the existence of a recommendation. We recognize that our
time cutoff is arbitrary, but the one-year window should provide a reasonable opportunity for the
lead underwriter to initiate coverage. As we discuss later, our results are robust to using six
month or two year windows.
We also collect data on Institutional Investor’s All-star Analyst Team. We match an IPO
to an all-star if the lead underwriter has an all-star (first-, second-, or third-team) in the same
industry as the issuer in the year of the issue or the prior year.9 To measure the quality of the
underwriter, we use Jay Ritter’s updated Carter-Manaster (1990) underwriter reputation
measures. We also use Ritter’s data to construct variables to measure whether an issue was
completed during a “hot market.” 10 Specifically, for each IPO, we measure market conditions in
two ways - as the total number of all IPOs (including those not in our sample) conducted during
the month of and the month prior to the IPO, and as the average underpricing across all IPOs
during the same two-month period. To get a firm-specific measure of a hot deal, we calculate a
turnover variable as the ratio of average daily volume over the thirty trading days following the
IPO to the number of shares issued.
C. Data description
Table I reports a time profile of the sample IPOs along with selected characteristics. The
number of offerings for which at least one SEO was conducted by the end of 2001 ranges from a
low of 38 in 2000 to a high of 210 in 1996.11 Consistent with the data reported in Ritter and
Welch (2002), average underpricing increases dramatically in the late 1990s. Although
underpricing averages 28% for the full sample, it averages 91% in 1999. Interestingly, although
the late-1990s exhibit the greatest underpricing, this period was not the most active period from
the point of view of number of deals, even before we apply our SEO requirement. In unreported
results we also find that the proportion of IPOs by technology companies in our sample was
12
much greater in the late 1990s than earlier (73% in 1999 vs. 31% in 1993). Columns four and
five of Table I show that the patterns of frequency and underpricing for our sample of IPOs are
representative of the overall population of IPOs issued during the same time period.
The sixth column of Table I shows the fraction of IPOs for which we can definitively
establish a link to the I/B/E/S database.12 It is clear that in the first two years of our sample there
are more unmatched deals. This means that we are potentially counting a deal as having no
coverage, when in fact there may be coverage that we were simply unable to identify. In Section
III.G., we show that our results are robust to excluding these deals. Overall, we match the lead
underwriter to I/B/E/S for 96% of our IPOs. Our match rates and coverage frequencies are
similar to those found in Krigman, Shaw, and Womack (2001).
Finally, in the last column we report the fraction of issuers who switch underwriters for
their SEO. We define an issuer as having switched if it does not employ the lead IPO
underwriter (or a subsequent affiliate through merger or acquisition) as the lead managing
underwriter in the first SEO following the IPO. An issuer that uses the IPO lead as a co-manager
or general syndicate member in the SEO is classified as switching. This definition of switching
is consistent with Krigman, Shaw, and Womack (2001). Our data indicate that 34% of issuing
firms switch the lead underwriter for their first SEO. Of the firms that switch lead underwriters,
approximately half employ the IPO lead underwriter as a co-manager in the SEO and half do not
employ the underwriter in the SEO at all. (These data are not reported in the table.) It is very
rare that the lead underwriter from the IPO is demoted to the position of a general syndicate
member in the SEO.
The rate of underwriter switching in our sample declines over time from 40% in 1993 to
16% in 2000. Of course, this pattern is likely due to the fact that (i) firms are more likely to
switch underwriters if there is a long time between their IPO and their SEO, and (ii) IPOs in the
early part of our sample potentially have a longer time period between IPO and SEO. We later
control for the length of this period in the logit regressions predicting the likelihood of switching
underwriters, and verify that the correlation between analyst coverage and switching
13
underwriters is similar if we limit the sample to those cases in which the firm completes its SEO
within three years of its IPO.
Table II reports summary statistics on a few other key variables. Across all IPOs,
underpricing ranges from a low of -29% to a high of 606%. The presence of some extreme
positive underpricing makes the median of 11.6% much less than the mean of 27.5%. The
average IPO uses an underwriter with a reputation measure of 7.5.13 About 22% of the issues
employ a lead underwriter who has an all-star in that industry. Issuers raised a mean of $66
million (in 2000 dollars), with a range from $2.5 million to $2.9 billion. As first documented by
Chen and Ritter (2000), the underwriting spread is clustered at 7%, with 74% of the IPOs having
a spread of exactly 7%. We also observe clustering at other integers such as 8% and 10% in our
sample. Forty-five percent of the sample firms are defined as technology companies, and 96%
are traded on a major market (e.g. NYSE, AMEX, or Nasdaq NMS). Finally, the average offer
price revision (i.e. the percentage difference between the offer price and the midpoint of the
filing range) is 3.1%, though the median IPO is issued at the midpoint of the filing range. We
observe large deviations in this variable, ranging from -60% to 140%.
III. Empirical Results
We begin our empirical analysis by reporting the frequency and distribution of post-IPO
analyst recommendations. We then examine the link between underpricing and analyst coverage
via univariate comparisons, ordinary least squares regressions, and two-stage OLS and logit
models that control for the endogeneity of underpricing and analyst coverage. Finally, we
examine whether the likelihood of switching underwriters for the company’s SEO is related to
the unexpected (conditional on underpricing) amount of post-IPO analyst coverage.
A. Analyst Coverage and Recommendations
Table III reports the extent of post-IPO analyst coverage and the strength of their
recommendations. The data in panel A indicate that most (75%) of the sample IPOs receive
14
coverage from a lead underwriter and at least one other analyst one year after the IPO date. Only
48 (4.6%) IPOs have coverage by the lead underwriter only.
Somewhat surprisingly, 117 offerings (11.1% of the sample) have no coverage by the
lead, but do have coverage by another analyst (which may include co-managers or other
syndicate members).14 The last two columns of the table provide some interesting information
about these deals. When the lead underwriter is the only bank providing coverage, the lead bank
tends to be of lower quality, as shown by an average reputation rank of 5.8 and 14.6% frequency
of all-stars. When the lead makes no recommendation but other banks do, the lead tends to be of
higher quality (7.4 reputation rank and 23.9% all-star frequency). These facts are consistent with
a situation in which underwriters value their reputation and lead underwriters would rather not
offend their clients by issuing unfavorable recommendations.
Finally, the last two rows present data on the IPOs for which there are no analyst
recommendations. There are 51 issuers for which we can determine a match between the SDC
and I/B/E/S databases, but for which there is no coverage by the lead or any other analyst. In
addition, there are 43 IPOs for which we are unable to definitively determine an SDC/ I/B/E/S
match. In all likelihood, most of these unmatched issuers probably do not get coverage as they
tend to be very small IPOs ($9.9 million average proceeds), in small industries, have low share
turnover, and are done by less prestigious underwriters (2.7 average reputation). These issuers
also are very likely (76%) to switch underwriters for the SEO. Our results are robust to the
exclusion of these 43 observations.
Panel B of Table III reports the frequency of different recommendations at the one-year
anniversary by lead and non-lead analysts. When there are multiple lead managers with
recommendations, we use the average recommendation, rounded to the nearest integer. Thus, for
example, if Strong Buy=5, Buy=4, and so on, and if there are two lead managers, one of whom
issues a buy recommendation (4) and one of whom issues a strong buy (5), this would average to
4.5. We would then round this to 5, a strong buy. Consistent with Bradley, Jordan, and Ritter
(2003), it is apparent that analysts either say something nice or say nothing at all. Analysts issue
15
no sell or strong sell recommendations and only 5.5% of the recommendations made by the lead
(3.2% of those made by the non-lead) are to Hold. Both leads and non-leads tend to split the
remaining recommendations fairly evenly between Strong Buy and Buy. For the issuers for
which both lead and non-lead underwriters make recommendations, the average recommendation
by a lead underwriter is a 4.49, versus a 4.37 for non-lead underwriter, where. This difference is
statistically significant at the 0.01 level (t = 4.7).
B. Univariate Comparisons of Underpricing and Analyst Coverage
In Panel A of Table IV, we first sort the sample IPOs into quintiles based on
underpricing, then compare average values of key variables across the quintiles. Some of these
data are also depicted graphically in Figure 1. Average underpricing ranges from –2.5% in the
lowest quintile to 98.7% in the highest quintile. Consistent with our hypothesis, analyst
coverage (recommendation or forecast) is positively related to underpricing. Ninety-four percent
of the firms in the highest quintile receive some coverage (recommendation or earnings
forecasts), as compared to about 85% in the lowest two quintiles. The pattern for lead
recommendations is similar, ranging from about 73% up to 86%. A test of equality across
quintiles rejects the hypothesis that underpricing is unrelated to analyst coverage at the 0.01
level.
These findings support the hypothesis that underwriters agree to provide coverage to
those issuers who agree to greater underpricing. However, consistent with Rajan and Servaes
(1997) and Krigman, Shaw, and Womack (2001), the next column shows that non-lead
underwriters are also more likely to cover deals that have large underpricing. Although the set of
non-lead underwriters includes co-managers who may also benefit from underpricing, this result
indicates that our subsequent tests will need to control for the possibility that greater
underpricing leads to greater coverage.
Consistent with Beatty and Welch (1996), there is a positive relation between
underpricing and the reputation of the underwriter. Similarly, the frequency of all-star coverage
16
roughly doubles as one moves from the lower three underpricing quintiles to the highest quintile.
Apparently, the issuers don't mind the underpricing. Consistent with the findings in Krigman,
Shaw, and Womack (2001), almost half of the low-underpricing firms switch underwriters, while
only a sixth of the high-underpricing firms switch. To the extent that highly underpriced IPOs
receive greater analyst coverage, this finding supports our hypothesis. However, another
explanation for this pattern, offered by Loughran and Ritter (2002a), is that the issuers with
greatest underpricing are happy because they ended up with greater proceeds (and wealth) than
they originally anticipated. Consistent with this view, we (like others) find a positive relation
between offer price revisions and underpricing. The least underpriced deals have a 12%
reduction from the midpoint of the filing range, whereas the most underpriced issues have a 26%
increase prior to the IPO. Finally, there is a strong industry effect in the underpricing quintiles.
Seventy-one percent of the IPOs in the highest quintile are technology firms, compared to about
35% to 45% for the other quintiles.
Panel B of the table repeats the exercise for many of the same variables, now splitting the
sample based on whether the lead underwriter makes a recommendation. When the lead makes a
recommendation, average underpricing is 30.5%, which is significantly larger than the average
of 15.7% when there is no lead recommendation. IPOs without lead coverage tend to be
underwritten by lower quality banks, have higher underwriting spreads, and have lower offer
price revisions. Consistent with our hypothesis, issuers who do not get a recommendation from
their lead IPO underwriter tend to be much more likely to use a different underwriter for their
first SEO (63% switch) than issuers who do get recommendations (26% switch).
C. Ordinary Least Squares Regression Results
To facilitate comparison of our results with the existing literature, we estimate ordinary
least squares (OLS) regressions in which underpricing is the dependent variable. Table V shows
three specifications, starting with one in which we do not include any analyst coverage-related
variables. All three models contain calendar year dummy variables to control for intertemporal
17
variation in average pricing. Consistent with our univariate findings, underpricing is positively
related to underwriter reputation and the offer price revision. The offer price revision variable is
a particularly strong determinant of IPO underpricing, consistent with the partial adjustment
phenomenon first reported in Hanley (1993).
We find weak evidence (t-statistics of about -1.7) of a negative relation between issue
size and underpricing, a significant negative relation for offerings not traded on a major
exchange, a significant positive relation for both the market-wide level of average IPO
underpricing and the CRSP value-weighted return, and a significant negative relation with firm
age.15 We find no relation to the frequency of IPOs in the market, the underwriter spread,
technology firms, or the volatility of market returns prior to the issuance. These findings are
generally consistent with those reported in the literature, providing further assurance that our
sample is representative of the population of issuing firms. Moreover, the regression model
explains a large portion of the cross-sectional variation in underpricing, as evidenced by the
adjusted R2 of 0.44.
To give some sense of the economic relevance of the significant coefficient estimates, an
increase in the underwriter reputation variable from a 7 (e.g. Legg Mason) to a 9 (e.g., Goldman
Sachs) is associated with an increase in underpricing of 4.5%. The point estimate of 0.89 on the
offer price revision variable indicates that as the offer price is revised up by 10% (say from $20
to $22), underpricing tends to rise by 8.9 percentage points.
In model (2) we add a dummy variable equal to one if the lead underwriter provides an
analyst recommendation. The inclusion of this variable essentially has no effect. The point
estimate is not significantly different from zero and is small in economic magnitude, the other
variables are not affected, and the adjusted R2 actually drops. This is inconsistent with our first
hypothesis which predicts a positive relation between underpricing and coverage. However, as
we demonstrate in the next section, it is important to control for the endogeneity between
underpricing and coverage.
18
Finally, in model (3) we add a dummy variable for the presence of an all-star analyst.
Consistent with our second hypothesis, this variable is both statistically and economically
significant. The point estimate indicates that underpricing is 9% higher in IPOs in which the
lead underwriter has an all-star analyst covering the industry of the IPO firm. This finding
supports the view that issuing companies value the presence of an all-star analyst and pay for this
prestige via underpricing. Most of the remaining coefficients are unaffected, although the role of
underwriter reputation is somewhat muted in the presence of the all-star dummy (almost all all-
stars are at banks rated 8 or 9).
D. Two-stage Estimation to Control for Endogeneity
One criticism of the OLS regressions in Table V is that they assume that analyst coverage
is exogenous. Based on the discussion in Section I, however, it is clear that underpricing and
analyst coverage may be endogenous. Similar to the approach adopted in Lowry and Shu
(2002), we attempt to mitigate the bias that this endogeneity induces in the regression
coefficients by using a two-stage estimation procedure. We estimate first-stage models of
underpricing and analyst coverage including the same set of exogenous variables in each
equation. Our choice of variables is motivated by the large literature on the determinants of
underpricing, as well as the determinants of analyst coverage. Specifically, we include variables
for the log of real proceeds, the lead underwriter's reputation, the relative size of the industry,
average trading volume for the thirty trading days following the IPO scaled by the number of
shares offered, the number of co-lead managers, the number of IPOs by any firm in the month of
the issue and the prior month, the average underpricing during this period, the gross underwriting
spread, the offer price revision, the average and standard deviation of returns on the value-
weighted CRSP index during the three weeks prior to the issuance, the log of one plus firm age,
and dummy variables for technology firms, all-star coverage by the lead underwriter, and
whether the firm is not listed on a major exchange. The underpricing regression is estimated by
19
OLS and the coverage model is estimated by logit. The coefficient estimates from these first-
stage models are reported in the first two columns of Table VI.
We then use the fitted values from these models as instruments in the second stage
estimation. The second stage models also include as independent variables those exogenous
variables that have a strong theoretical justification. The standard errors for the second-stage
estimates correct for estimation error in the first stage using the procedure described in Maddala
(1983).
The results in the third column of Table VI identify two main determinants of coverage.
The first is the reputation of the lead underwriter, which is positive and highly significant (t =
6.0). To interpret the economic magnitude, we compare the estimated probability of coverage at
the sample mean, where the underwriter reputation is 7.5, to the probability when the reputation
rank increases to the maximum of 9. Our estimates indicate that moving from an average
underwriter to the most reputable underwriter increases the likelihood of coverage by 6.5%. The
all-star variable is negative and significant, with a t-statistic of –2.2. Again, we evaluate the
economic impact of moving from no all-star to having an all-star. The impact of the all-star is a
drop in the likelihood of coverage of 8.2%. This comparative static is somewhat misleading
since it is unlikely that a firm would have an underwriter with average reputation and an all-star.
When we combine these two effects, they largely offset. In comparing an issuer using an
average reputation underwriter with no all-star to an otherwise identical issuer using a highly
reputable underwriter with an all-star, the likelihood of coverage drops by 0.4%. Finally, we
note that the underpricing instrument is positive, but not significantly different from zero.
Overall, the model has a pseudo-R2 of 0.173, correctly classifying 84.9% of the IPOs.
The last column of Table VI shows the results for the underpricing regression.
Consistent with our second hypothesis, we find that the presence of an all-star analyst increases
underpricing by an economically large 13.9 percentage points (t-statistic of 3.6). However,
partially offsetting this effect, a one-point increase in the underwriter's rank lowers underpricing
by 1.52 percentage points. In comparing an issuer with an underwriter of average reputation
20
(7.5) and no all-star analyst to an identical issuer with a highly reputable underwriter (9) and an
all-star analyst, we find that underpricing is increased in the second case by 11.6 percentage
points.
We also observe a strong positive relation between the spread and underpricing (t = 2.8).
Increasing the spread by a percentage point increases underpricing by 11 percent. As other
researchers have shown, the offer price revision is a strong predictor of underpricing (t = 9.7).
Given the point estimate of 0.78, a one standard deviation increase in the revision raises
underpricing by 17.4%.16 Underpricing is related to pre-issuance conditions in the IPO market.
Underpricing is higher when average underpricing across all recent IPOs is high (t = 5.9) and,
consistent with Benveniste et al. (2003), lower when the volume of IPOs is high (t = -2.7).17
Underpricing is also positively related to the pre-issuance value-weighted market return (t = 2.2).
Old firms have lower underpricing than young firms (t = -2.4), consistent with the notion that
underpricing is related to uncertainty about the issuer. We also find evidence that technology
firms have greater underpricing after controlling for other determinants of underpricing.
Of primary interest is the coefficient on the instrument for analyst coverage. Consistent
with our hypothesis, we find a strong positive relation between the coverage instrument and
underpricing (t = 3.2). Unfortunately, it is not possible to determine the economic impact of
expected analyst coverage on underpricing since the unidentifiable volatility of residuals in the
first-stage logit introduces a nuisance parameter. Overall, the regression has an adjusted R2 of
0.45. These findings support the view that the likelihood of subsequent analyst coverage is an
important determinant of the magnitude of underpricing. One interpretation of this finding is
that issuing companies pay for expected analyst coverage by discounting the price at which they
sell new shares.
We caution the reader that because some of the exogenous variables that predict
underpricing also predict analyst coverage, part of their impact on underpricing may be picked
up by the coverage instrument. If so, collinearity with the coverage instrument will increase the
standard errors of the coefficient estimates. One should, therefore, interpret the magnitude and
21
statistical significance of the coefficients on the exogenous variables with caution. We note,
however, that the coefficient estimates are, with the exception of underwriter rank, similar in
sign and statistical significance to those reported for the OLS regressions in Table V. This
provides some reassurance that our findings are not driven by our instrumental variables
approach. Nonetheless, it should be noted that the significance of the coverage instrument is
sensitive to the inclusion of year dummies in the second stage models. Because we attempt to
capture time trends in the data by including year dummies in the first stage, inclusion of the year
dummies in the second stage induces fairly severe collinearity problems. This shows up in the
form of substantially larger standard errors on the coefficient estimates after having made the
adjustment for first-stage estimation. Consequently, virtually nothing is statistically significant if
we include the year dummies in the second stage.
E. Subperiod Results
Because the 1998-2000 period exhibits dramatically higher underpricing and Loughran
and Ritter (2002b) document nonstationarities in some of the cross-sectional determinants of
underpricing, we also estimate the models in Table VI for three separate subperiods: 1993-1994,
1995-1997, and 1998-2000. The first subperiod represents the period in which we are less able
to link the SDC data with the I/B/E/S data, thereby raising the possibility that we incorrectly
conclude that the issuing firm receives no coverage. The third subperiod represents the period of
unusually high underpricing as well as greatly increased analyst coverage.
In Panel A of Table VII, we report descriptive statistics for the three subperiods. Not
surprisingly, average underpricing is approximately four times larger in the 1998-2000 subperiod
than in the 1995-1997 period. Perhaps more interestingly, the 1998-2000 period also exhibits a
large increase in the percentage of issuing companies that choose a lead underwriter with an all-
star analyst (39.9% vs. 18.2%), but little difference in the frequency with which the lead
underwriter provides analyst coverage (87.8% vs. 86.3%).
22
In Panel B, we report selected coefficients from two-stage underpricing regressions
identical to those estimated in Table VI. We note at the outset that these coefficients should be
interpreted with caution due to the smaller sample sizes. For example, because there are only 31
issues that do not receive analyst coverage in the 1998-2000 period, the power of the test of the
coverage instrument in these models is fairly low. Nonetheless, the analysis yields some
interesting results. Although we observe little change in the coefficient on the coverage
instrument, the coefficient on all-star analyst in the first-stage underpricing regression is
substantially larger in the third subperiod than in the second subperiod (21.13 vs 0.43). This is
also true in the second stage regressions (15.14 vs. 2.72), but the coefficients lack statistical
significance.
Subject to the caveat noted above, these findings are broadly consistent with Loughran
and Ritter’s (2002b) analyst lust hypothesis. It appears that in the latter part of the 1990s, issuing
companies (i) exhibited a stronger demand for all-star analyst coverage and (ii) were willing to
give up greater underpricing for this coverage. Both effects potentially contribute to the large
increase in underpricing in the 1998-2000 period, though they are clearly not large enough to be
the only explanation.
It is also noteworthy that the coefficient on the coverage instrument is significant in both
of the first two subperiods. This provides some reassurance that our overall finding of a
significant relation between underpricing and coverage is not driven by the 1998-2000 period.
F. Switching of Underwriters
Our final hypothesis predicts that issuing companies will switch underwriters between
their IPO and their subsequent SEO if they believe they have received less analyst coverage than
expected. To test this hypothesis, we examine how coverage and underpricing jointly affect an
issuer's decision to switch underwriters at the SEO.
Recall from Table IV that there is an inverse relationship between underpricing and the
likelihood of switching underwriters. To further address why the issuers leaving the most money
23
on the table are the least likely to switch underwriters, Table VIII compares the switching rates
in underpricing quintiles of firms with and without lead analyst recommendations. Within a
given underpricing quintile, firms that get lead coverage are much less likely to switch. For
example, in the low underpricing quintile, where issuers are very likely to switch underwriters,
74% of the issuers who do not get coverage switch, as compared to a 37% switching rate among
the issuers who receive lead coverage. The other quintiles exhibit a similar pattern, with the
switching rate of firms with lead analyst coverage being roughly 30 percentage points below that
of firms without analyst coverage. For all five quintiles, the difference in the percentage of firms
switching underwriters between those with a lead analyst recommendation and those without
such a recommendation is significant at the 1% level.
On the other hand, splitting issuers into coverage categories does not remove the spread
across underpricing quintiles. For firms with recommendations from the lead underwriter, the
37% switch rate for the low-underpricing quintile is three times that of the high-underpricing
quintile. Similarly, among firms without recommendations from the lead underwriter, the 74%
switching rate in the low-underpricing quintile is nearly double the rate for the high-underpricing
quintile. These findings suggest that analyst coverage is only part of the explanation for why
issuing firms switch underwriters.
To provide further evidence on the determinants of underwriter switching, we estimate
logit models to predict switching behavior. Our analysis is similar to that in Krigman, Shaw, and
Womack (2001), with one important addition. We include in our model the unexpected analyst
coverage (actual coverage minus the predicted probability) from our second-stage estimates in
Table VI. The results are reported in Table IX.18
We consider a base model using a constant, the log of offer proceeds, offer price revision,
share turnover, underwriter spread, dummy for an all-star analyst at the IPO and SEO lead
underwriter, IPO and SEO underwriter rank, the number of calendar days from IPO to SEO, the
log of one plus firm age, and IPO underpricing. We find that switching is more likely for firms
that have a small offer price revision, firms whose IPO underwriter has a lower reputation, firms
24
whose SEO underwriter has a high reputation, and firms for which there is a long time between
IPO and SEO.
The economic impact of changes in the explanatory variables is shown in the third
column. From this analysis, it is clear that the reputation of the underwriter is a primary
determinant of the likelihood of switching. A one standard deviation increase in the rank of the
IPO underwriter reduces the probability of switching by 20 percent. Similarly, a one standard
deviation increase in the reputation of the SEO underwriter increases the likelihood of switching
by 19 percent. These findings are consistent with the graduation story in Krigman, Shaw, and
Womack (2001). Firms appear to gravitate towards the more reputable underwriters for their
SEO if they used a less prestigious underwriter for their IPO. The chance of switching is also
reduced by the offer price revision, perhaps because these issuers tend to be pleased that they
raised more funds than they originally anticipated. Increasing the offer price revision by one
standard deviation reduces the chances of switching by 7 percent. Finally, a one standard
deviation change in the number of days between IPO and SEO increases the likelihood of
switching by 20 percent. It seems plausible that the strength of the relationship between
underwriters and issuers would decay over time.
The last set of columns in Table IX augment the base model with a measure of
unexpected coverage. Our third hypothesis predicts that if a firm receives less coverage than
expected, they will be more likely to use a different underwriter for their SEO. We find that this
is indeed the case. The unexpected coverage variable has a t-statistic of –4.8. Unfortunately, we
are unable to assess the economic significance for the same reason as in Table VI.19
G. Robustness Checks
To ensure that our results are not driven by methodological choices or a small number of
influential observations, we run a battery of robustness checks. One group of tests replicates all
our analyses after filtering the sample in a variety of ways. First, we exclude the 160
observations for which the IPO was completed in 1999 or 2000. This addresses the concern that
25
our findings are biased by the fact that firms completing their IPO in these years have done SEOs
quickly, relative to the rest of the sample. Truncating the sample in 1998 allows each firm three
years to complete an SEO, which is approximately double the average of 1.55 years between IPO
and SEO for firms in this subsample. Second, we exclude firms with offer prices below $8, as in
Loughran and Ritter (2002a). This reduces our sample to 920 firms. Third, we exclude
observations in the extreme 1% tails of the underpricing distribution. Fourth, we exclude the 111
observations for which the company’s SEO takes place more than three years after the IPO.
Fifth, because I/B/E/S’s coverage of analyst recommendations may have been less complete
prior to 1995, we exclude 354 offerings completed in 1993 and 1994.20 Sixth, we restrict the
sample to include only IPOs completed after 1994 and those for which the company’s SEO takes
place more than three years after the IPO. This reduces the sample by 402 observations.
Seventh, we restrict the sample to only those firms that initially trade on the NYSE, AMEX, or
Nasdaq NMS. In all cases, our main results are not affected in any material way. Specifically,
we continue to find a positive relation between underpricing and predicted coverage, and
continue to find that the likelihood of switching underwriters at the time of the SEO is negatively
related to unexpected analyst coverage following the IPO.
The second group of robustness test focuses on methodological choices. Again, none of
these checks meaningfully alters our main results. First, we estimate all logit models by probit.
Second, we delete from our main sample the 43 offers for which we are unable to link SDC
underwriters with I/B/E/S brokers. Our main analysis considers these IPOs as having received
no coverage. However, it is possible that these deals do get coverage but either I/B/E/S does not
follow that brokerage firm or we did not properly identify the link between SDC and I/B/E/S
bank codes. Third, we exclude observations for which the time between IPO and SEO is less
than one year. Recall that we measure coverage as of one year after the issuance, so for these
deals we are measuring coverage after the SEO. This results in a loss of about half our sample,
down to 518 firms, of which 370 have coverage. This sub-sample has much lower underpricing
(13% on average) and much higher switching rates for the SEO underwriter (50% on average).
26
However, our main results remain intact. Underpricing is positively associated with expected
coverage, while the likelihood of switching underwriters is negatively related to unexpected
coverage. These findings also indicate that our primary results are not driven by successful
companies that quickly issue an SEO in the first year following their IPO. Fourth, we include
the annualized stock returns between IPO and SEO as an explanatory variable. Again, our
results are unaffected.
A third group of robustness checks reconstructs the sample using alternative windows for
measuring analyst coverage. First, we record a firm as receiving recommendation coverage if it
has a recommendation from the lead underwriter six months after the IPO. This increases the
number of firms without coverage from 237 to 291. Our main results remain intact. Second, we
repeat the analysis after measuring analyst coverage as of the two-year anniversary of the IPO.
Because this means we are checking for coverage well after many firms have done at least one
SEO, we again filter out deals where there is less than a year between IPO and SEO. Although
this reduces the sample to 518 observations, of which 350 have coverage, our main results are
robust. Finally, we measure coverage as receiving a recommendation during any point in the
first year following the IPO. By this measure, a firm that receives coverage for only a few
months is counted as receiving coverage. This less restrictive measure records 874 deals with
lead coverage, compared to 839 in the main sample, but does not change our results.
Finally, we examine the possibility that lead underwriters choose not to provide
recommendations on some firms because they deem these particular issuers to be sufficiently
unimportant to merit any analyst coverage. To examine this issue, we first create a sub-sample
of IPOs for which the lead underwriter provides earnings forecasts. We know for sure that the
analyst is following these firms. We then split these firms into two groups based on whether the
analyst of the lead underwriter also makes a recommendation. Of the 928 firms with earnings
forecasts from the lead underwriter, 830 also have a lead recommendation and 98 do not.21
Those issuers receiving recommendations have average underpricing of 30%, significantly
greater than the 19% average for those who do not have recommendations. In addition, we
27
observe that among those firms that do not receive a lead recommendation, 55% switch
underwriters for their SEO. This happens in only 26% of the cases in which there is a lead
recommendation. Thus, among the subset of firms for which the lead underwriter provides
analyst coverage, (i) underpricing is significantly greater for firms receiving analyst
recommendations, and (ii) firms are significantly more likely to switch underwriters if the lead
IPO underwriter chooses not to issue a recommendation. The fact that our main results continue
to hold for the sub-sample of firms that clearly receive some analyst attention provides
reassurance that our main findings are not driven by cases in which the analyst of the lead
underwriter simply ignores issuers that they deem to be unimportant. Our results are more
consistent with the view that the lack of a recommendation is driven by strategic considerations.
That is, banks seek to avoid offending their clients by making negative recommendations, but
also want to avoid ruining their reputations by providing favorable coverage to issuers with poor
prospects.
IV. Discussion and Concluding Remarks
We examine the links among IPO underpricing, post-IPO analyst coverage, and the
likelihood of switching underwriters. Our findings indicate a significant positive relation
between underpricing and analyst coverage by the lead underwriter. This positive association is
robust to controls for other determinants of underpricing previously documented in the literature
and to controls for the endogeneity of underpricing and analyst coverage. In addition, after
controlling for other potential determinants of switching underwriters, we find that the
probability of switching underwriters between IPO and SEO is negatively related to the
unexpected amount of post-IPO analyst coverage. We interpret these findings as consistent with
the hypothesis that underpricing is, in part, compensation for expected post-IPO analyst
coverage. If underwriters do not deliver the expected analyst coverage (conditional on
underpricing), the IPO firm is more likely to switch underwriters when it issues shares in its
subsequent SEO.
28
An alternative explanation for the positive correlation between underpricing and analyst
coverage is that issuers deliberately underprice IPOs in order to attract analyst attention and build
price momentum for open market sales following the expiration of the lockup period [Aggarwal,
Krigman, and Womack (2002)]. While this strategic underpricing explanation and our
hypothesis are not necessarily mutually exclusive, some of our findings are difficult to reconcile
with strategic underpricing. Specifically, it is not clear why there would be any connection
between analyst coverage and the likelihood of switching underwriters. Moreover, under the
strategic underpricing hypothesis, it is less clear why underpricing should be higher in deals
underwritten by investment banks with an all-star analyst.
Our findings can help explain a few otherwise puzzling IPO phenomena. First, recent
studies [e.g., Beatty and Welch (1996)] report that the correlation between underpricing and
underwriter reputation has changed signs from negative in the 1970s and 1980s [Carter and
Manaster (1990)] to positive in the 1990s. To the extent that analyst coverage has become more
important in the past decade, as argued in Loughran and Ritter (2002b), our hypothesis predicts
that more prestigious underwriters will be compensated for expected analyst coverage with
greater underpricing.
Second, the increased importance of analyst coverage in recent years can help explain the
large increase in the salaries of sell-side analysts during the late 1990s. Our hypothesis predicts
that investment banks receive additional compensation, via underpricing, for the research
coverage that they provide. Presumably, a portion of this compensation is passed on to the
analysts providing such coverage. Of course, as underwriting business and merger/acquisition
activity has declined over the past couple of years, so too has analyst compensation. This has led
to some high profile departures of analysts and to large cutbacks in the research staff at Wall
Street firms.22
Finally, our findings suggest a possible reason why issuing companies do not appear to
be upset by the underpricing of their IPOs. If underpricing is, in part, compensation for
subsequent research coverage, issuers might be getting exactly what they pay for, on average. Of
29
course, as Loughran and Ritter (2002b) argue, underpricing may still be too large, thereby
leading to excessive underwriter compensation. Our findings are silent on this issue.
30
References
Aggarwal, R., L. Krigman, and K. Womack, 2002, Strategic IPO Underpricing, Information
Momentum, and Lockup Expiration Selling, Journal of Financial Economics 66, 105-
137.
Beatty R.P. and Welch, I. 1996, Issuer Expenses and Leagal Liability in Initial Public Offerings,
Journal of Law and Economics 39 (2), 545-603.
Benveniste, L.M., A.P. Ljungqvist, W.J. Wilhelm, and X. Yu, 2003, Evidence of Information
Spillovers in the Production of Investment Banking Services, Journal of Finance 58, 577-
608.
Bradley, D., B. Jordan, and J.R. Ritter, 2003, The Quiet Period Goes Out With a Bang, Journal
of Finance 58, 1-36.
Carter, R.B. and S. Manaster, 1990, Initial Public Offerings and Underwriter Reputation, Journal
of Finance 45, 1045-1067.
Chemmanur, T., 1993, The Pricing of Initial Public Offerings: A Dynamic Model with
Information Production, Journal of Finance 48, 285-304.
Chen, H. and J.R. Ritter, 2000, The Seven Percent Solution, Journal of Finance 55, 1105-1131.
Clarke, J., C. Dunbar, and K. Kahle, 2003, All-Star Analyst Turnover, Investment Bank Market
Share, and the Performance of Initial Public Offerings, Working Paper, University of
Pittsburgh.
Das, S., R. Guo, and H. Zhang, 2002, Initiation of Analyst Coverage and Long-term Performance
of Initial Public Offerings, Working Paper, University of Illinois-Chicago.
Dunbar, C.G., 2000, Factors Affecting Investment Bank Initial Public Offering Market Share,
Journal of Financial Economics 55, 3-41.
31
Ellis, K., R. Michaely, and M. O’Hara, 2000, When the Underwriter is the Market Maker: An
Examination of Trading in the IPO Aftermarket, Journal of Finance 55, 1039-1074.
Field, L. and J. Karpoff, Takeover Defenses at IPO Firms, Journal of Finance 57, 1857-1889.
Grullon, G., G. Kanatas, and J. Weston, 2003, Advertising, Breadth of Ownership, and Liquidity,
Review of Financial Studies, forthcoming.
Habib, M.A. and A.P. Ljungqvist, 1998, Underpricing and IPO Proceeds: a note, Economics
Letters 61, 381-383.
Hakenes, H., and Nevries, P., 2000, Underpricing Initial Public Offerings Due to the Value
Increasing Publicity Effect, Working Paper, Westfalishce Wilhelms-Universitat Munster.
Hanley, K.W., 1993, The Underwriting of Initial Public Offerings and the Partial Adjustment
Phenomenon, Journal of Financial Economics 34, 231-250.
Hansen, R.S., 2001, Do Investment Banks Compete in IPOs? The Advent of the “7% plus”
Contract, Journal of Financial Economics 59, 313-346.
Jegadeesh, N., J. Kim, S. Krische, and C. M. C. Lee, 2002, Analyzing the Analysts: When do
Recommendations Add Value? Working Paper, University of Illinois.
Krigman, L., W.H. Shaw, and K.L. Womack, 2001, Why Do Firms Switch Underwriters?
Journal of Financial Economics 60, 245-284.
Lin, H. and M.F. McNichols, 1998, Underwriting Relationships, Analysts’ Earnings Forecasts,
and Investment Recommendations, Journal of Accounting and Economics 25, 101-127.
Ljungqvist, A. and W.J. Wilhelm, 2002, IPO Allocations: Discriminatory or Discretionary?
Journal of Financial Economics 65, 167-201.
Loughran, T. and J.R. Ritter, 2002a, Why Don’t Issuers Get Upset About Leaving Money on the
Table in IPOs? Review of Financial Studies 15, 413-443.
32
Loughran, T. and J.R. Ritter, 2002b, Why Has IPO Underpricing Changed Over Time? Working
Paper, University of Florida.
Lowry, M. and S. Shu, 2002, Litigation Risk and IPO Underpricing, Journal of Financial
Economics 65, 309-336
Maddala, G.S., 1983, Limited-dependent and Qualitative Variables in Econometrics (Cambridge
University Press, Cambridge).
Merton, R.C., 1987, A Simple Model of Capital Market Equilibrium With Incomplete
Information, Journal of Finance 42, 483-510.
Michaely, R. and K.L. Womack, 1999, Conflict of Interest and the Credibility of Underwriter
Analyst Recommendations, Review of Financial Studies 12, 653-686.
Mikhail, M.B., P. Asquith, and A. Au, 2002, Information Content of Equity Analyst Reports,
Working Paper, MIT.
Murphy, K. M. and R. H. Topel, 1985, Estimation and Inference in Two-Step Econometric
Models, Journal of Business and Economic Statistics 3, 370-379.
Rajan, R.G. and Servaes, H., 1997, Analyst Following of Initial Public Offerings, Journal of
Finance 52, 507-529.
Ritter, J.R. and Welch, I., 2002, A Review of IPO Activity, Pricing and Allocations, Journal of
Finance 57, 1795-1828.
Womack, K.L., 1996, Do Brokerage Analysts Recommendations Have Investment Value?
Journal of Finance 51, 137-167.
33
Appendix Construction of Variables
Variable Data Sources Description Underpricing SDC, CRSP Percentage return from offer price (SDC) to first day
close (CRSP) IPO frequency Ritter Number of IPOs in month of issue and prior month IPO returns Ritter Average IPO underpricing in month of issue and prior
month Underwriter Rank Ritter 1 (worst) to 9 (best) scale for underwriter reputation All-star dummy Institutional
Investor 1 if lead underwriter has an all-star in issuer’s industry during year of IPO or prior year
Proceeds SDC, Bureau of Labor Statistics
Offer proceeds (SDC) converted to 2000 dollars based on CPI from BLS.
Underwriter spread SDC Gross underwriter spread, in percent Tech dummy SDC 1 if issuer is a technology firm (SICs 2833, 2834,
2835, 2836, 3571, 3572, 3575, 3577, 3578, 3661, 3663, 3669, 3674, 3812, 3823, 3825, 3826, 3827, 3829, 3841, 3845, 4812, 4813, 4899, 7370, 7371, 7372, 7373, 7374, 7375, 7377, 7378, 7379)
Offer price revision SDC Percentage difference between offer price and midpoint of filing range
Non-exchange traded
SDC 1 if exchange is not NYSE, AMEX, or Nasdaq NMS
Recommendation dummy
I/B/E/S 1 if the lead has a recommendation for the issuer one year post-IPO. With joint managers, a 1 if any manager has a recommendation.
Recommendation level
I/B/E/S Recommendation (5=Strong Buy, 1 = Strong Sell) made by lead 1-year post IPO. Average if there are joint managers.
Industry Size CRSP Market cap of 3-digit SIC as a percentage of the total market cap on CRSP, computed annually.
Share turnover SDC, CRSP Avg. trading volume first 30 trading days post-IPO (CRSP), divided by shares issued (SDC).
# of co-lead managers SDC Number of co-managers (including lead manager(s)) Age Ritter, Field Year of IPO minus founding year. Most observations
from Ritter, with 32 missing observations augmented from other sources (Business and Company Resource Center Database, 10-K reports)
Pre-IPO Mkt Return CRSP Average return on CRSP value-weighted index from 3 weeks pre-issuance to issuance date.
Pre-IPO Mkt Std Dev CRSP Standard deviation of returns on CRSP value-weighted index from 3 weeks pre-issuance to issuance date.
34
Table I
Time Profile Time profile and selected characteristics of a sample of 1,050 initial public offerings (IPOs) completed between 1993 and 2000. Underpricing is measured as the percentage return from the offer price to the closing price on the first day of trading. We define a firm as having an I/B/E/S SDC link if we are able to match the lead underwriter of the IPO from SDC with an investment bank listed on I/B/E/S. The IPOs in the sample all complete a subsequent seasoned equity offering (SEO) between 1993 and 2001.
Year
# of IPOs
Average Underpricing
(%)
Average frequency of IPOs in current or
prior month
Average underpricing of IPOs in current or
prior month
Percent with
an I/B/E/S/SDC
Link
Percent that switch lead underwriter
at SEO
1993 191 13.0 108.0 15.8 93.7 40.3 1994 163 9.5 99.5 14.0 85.3 48.5 1995 155 18.2 102.0 20.0 98.7 31.0 1996 210 17.8 147.7 17.9 99.0 33.8 1997 108 16.7 104.9 14.6 97.2 33.3 1998 63 48.0 71.8 21.2 100.0 20.6 1999 122 91.2 89.9 65.3 100.0 18.0 2000 38 61.0 76.3 52.7 100.0 15.8 All 1050 27.5 108.0 23.8 95.9 33.5
35
Table II
Descriptive Statistics for IPOs Summary measures for a variety of sample characteristics. The sample includes 1,050 initial public offerings (IPOs) completed between 1993 and 2000 for which a subsequent seasoned equity offering (SEO) is made between 1993 and 2001. Underwriter rank is based on Jay Ritter’s updated Carter-Manaster (1990) measure.
Characteristic Mean Median Minimum Maximum Underpricing (%) 27.5 11.6 -29.2 605.6 Underwriter rank 7.5 8.0 1.0 9.0 Percent with all-star analyst 22.4 n.m. n.m. n.m. Percent with analyst forecast or recommendation at one-year anniversary of IPO
89.2 n.m n.m. n.m.
Percent with analyst recommendation at one-year anniversary of IPO
79.9 n.m. n.m. n.m.
Proceeds (in $millions) 65.5 41.0 2.5 2,853.1 Underwriter spread 7.1 7.0 4.0 10.2 % of offerings with non-7% spread
25.6 n.m. n.m. n.m.
Percent technology companies 44.9 n.m. n.m. n.m. Offer price revison between filing and offering (%)
3.1 0.0 -60.0 140.0
Percent not listed on organized exchange
3.8 n.m. n.m. n.m.
Age of company 11.6 6.0 0.0 145.0 n.m. – not meaningful
36
Table III
Analyst Coverage and Recommendations Frequency of analyst coverage and nature of recommendations one year after the IPO. The sample includes 1,050 initial public offerings (IPOs) completed between 1993 and 2000 for which a subsequent seasoned equity offering (SEO) is made between 1993 and 2001. For each offering we identify whether the offering company is covered either by the lead underwriter(s), non-lead underwriters, or neither, according to the Institutional Brokers Estimate System (I/B/E/S). For multiple recommendations from joint lead underwriters, the average is used, with rounding to the nearest integer.
Panel A: Frequency of Coverage
Number % of total Mean
Underwriter Rank
Percent with All-star
Lead and Non-lead Underwriter 791 75.3 8.0 24.7 Lead Underwriter Only 48 4.6 5.8 14.6 Non-lead Underwriter Only 117 11.1 7.4 23.9 Neither Lead nor Non-Lead 51 4.9 4.8 9.8 Unable to link I/B/E/S with SDC 43 4.1 2.7
Panel B: Distribution of Recommendations Lead Underwriters Non-Lead Underwriters
Number Percent of
recommendations
Number Percent of
recommendationsStrong buy (5) 455 54.2 454 44.8 Buy (4) 338 40.3 424 52.0 Hold (3) 46 5.5 40 3.2 No Recommendation 211 132 Average recommendationa 4.49 4.37
t-test of difference (p-value)
4.74 (0.0000)
a Includes only those IPOs in which both the lead and non-lead make a recommendation.
37
Table IV IPO Characteristics Sorted Based on Underpricing and Lead Coverage Status
Average characteristics of analyst coverage, underwriter characteristics, underwriter fees, the propensity to switch underwriters at the time of an SEO, the price revision between IPO offering filing and offering date, and the fraction of technology firms by quintile of IPO underpricing. The sample includes 1,050 initial public offerings (IPOs) completed between 1993 and 2000 for which a subsequent seasoned equity offering (SEO) is made between 1993 and 2001. p-values are reported for the significance of a test of equal means values across quintiles. KW p-values are Kruskal-Wallis p-values for tests of equal medians.
Panel A: Underpricing Quintiles
Under-pricing Quintile
Under-pricing
(%)
Percent with lead analyst forecast or
recommendation
Percent with lead analyst
recommend.
Percent with a non-
lead recommend.
Under-writer rank
Percent with an all-star analyst
Percent that switch underwriter
at SEO
Under- writer spread
Offer Price
revision(%)
Percent in technology
industry Low -2.5 87.1 74.8 81.4 7.2 15.7 46.2 7.1 -12.0 45.2 Q2 3.8 83.5 71.2 82.1 7.1 16.0 44.8 7.1 -8.0 37.7 Q3 12.1 92.3 82.7 86.5 7.3 18.3 33.7 7.0 0.9 33.7 Q4 25.4 89.0 85.2 91.0 7.7 26.7 26.2 7.1 8.7 36.2 High 98.7 94.3 85.7 96.7 8.2 35.2 16.7 7.1 26.0 71.4 p-value 0.0000 0.0031 0.0001 0.0000 0.0000 0.0000 0.0000 0.8195 0.0000 0.0000 KW p-value
0.0000 0.0032 0.0002 0.0000 0.0000 0.0000 0.0000 0.9128 0.0000 0.0000
Panel B: Recommendations by Lead Underwriter
Recommendations by Lead Underwriter
Under-pricing
(%)
Underwriter rank
Percent with
an all-star analyst
Percent that switch
underwriter at SEO
Underwriter spread
Offer Price
revision (%)
Percent in technology
industry No 15.7 5.8 15.6 62.6 7.7 -1.9 42.7 Yes 30.5 7.9 24.1 26.2 6.9 4.4 45.4 p-value 0.0002 0.0000 0.0086 0.0000 0.0000 0.0003 0.4721 KW p-value 0.0000 0.0000 0.0086 0.0000 0.0000 0.0002 0.4718
Table V OLS Regression ResultsWith Underpricing as the Dependent Variable
Cross-sectional regressions of percentage IPO underpricing on calendar year dummy variables (not reported), the log of real proceeds in year 2000 dollars, underwriter rank, the frequency of IPOs in the market during the current or prior month, the average underpricing of IPOs over the current or prior month, the underwriter spread, the price revision between the midpoint of the initial filing range and the offer price, a dummy variable for offerings not listed on NYSE, AMEX, or NASDAQ NMS, a dummy variable for technology companies, the average CRSP value-weighted index return over the three weeks up to issuance, the standard deviation of CRSP value-weighted index return over the three weeks up to issuance, the log of one plus firm age at issuance, a dummy variable equal to one if the lead underwriter makes a recommendation, and a dummy variable equal to one of the lead underwriter has an All-star analyst covering the industry of the IPO company. Coefficients are reported with heteroskedasticity-consistent t-statistics in parentheses below. The sample includes 1,050 initial public offerings (IPOs) completed between 1993 and 2000 for which a subsequent seasoned equity offering (SEO) is made between 1993 and 2001.
Variable Model (1) Model (2) Model (3) Log (proceeds) -3.88
(-1.68) -3.69
(-1.59) -4.14
(-1.77) Underwriter rank 2.25
(3.50) 2.07
(3.28) 1.45
(2.26) IPO frequency -0.02
(-0.39) -0.03
(-0.41) -0.03
(-0.44) IPO returns 0.58
(2.03) 0.58
(2.02) 0.60
(2.07) Underwriter spread 3.41
(1.60) 3.84
(1.75) 3.33
(1.52) Offer price revision 0.89
(8.42) 0.89
(8.32) 0.88
(8.08) Non-exchange traded
-7.41 (-2.17)
-7.28 (-2.14)
-7.88 (-2.30)
Technology dummy 4.14 (1.51)
4.27 (1.55)
3.79 (1.41)
Pre-IPO Mkt Ret 0.25 (2.95)
0.25 (2.96)
0.25 (2.92)
Pre-IPO Mkt Std 0.04 (0.64)
0.04 (0.64)
0.04 (0.70)
Log(1+Age) -1.76 (-1.95)
-1.85 (-2.01)
-1.78 (-1.94)
Lead underwriter recommendation
3.01 (1.25)
3.41 (1.38)
All-star analyst 8.73 (2.18)
Year dummies Yes Yes Yes Adjusted R2 0.440 0.440 0.444
39
Table VI Two-Stage Regression Results
Results of two-stage estimation of coverage and underpricing equations to control for endogeneity. Coverage equations are estimated by logit and underpricing is estimated by OLS. In the coverage equations, the dependent variable is equal to one if the lead underwriter makes a recommendation as of the one-year anniversary of the IPO. First stages estimates include all exogenous variables. Second stage estimates include subsets of exogenous variables, plus the fitted instrument (X’β) from the first stage regressions. Coefficients are reported with t-statistics in parentheses below. t-statistics from the second stage account for estimation error in the first stage following Maddala (1983). The sample includes 1,050 initial public offerings (IPOs) completed between 1993 and 2000 for which a subsequent seasoned equity offering (SEO) is made between 1993 and 2001.
First Stage Second Stage
Variable Coverage
Logit Underpricing
OLS Coverage
Logit Underpricing
OLS Constant 7.16 -73.77 -5.69 -39.44 ( 1.61) (-1.45) (-2.04) (-0.59) Log (proceeds) -0.26 2.18 0.24 -1.45 (-1.27) ( 0.91) ( 1.35) (-0.49) Technology dummy -0.31 0.90 -0.11 7.48 (-1.09) ( 0.27) (-0.42) ( 2.40) Underwriter rank 0.33 0.51 0.38 -1.52 ( 5.07) ( 0.85) ( 6.03) (-1.05) All-star analyst -0.31 9.01 -0.54 13.92 (-1.26) ( 2.49) (-2.17) ( 3.64) Non-exchange traded -0.11 -6.99 -0.53 (-0.23) (-2.31) (-1.11) Industry Size -0.02 0.28 -0.00 (-0.39) ( 0.29) (-0.09) Share turnover 0.02 1.64 0.01 ( 1.10) ( 2.56) ( 0.41) # of co-lead managers 0.11 -3.09 0.12 ( 0.73) (-1.77) ( 0.96) IPO frequency 0.00 0.00 -0.15 ( 1.23) ( 0.09) (-2.69) IPO returns 0.02 0.48 0.70 ( 1.57) ( 1.89) ( 5.91) Underwriter spread -0.78 4.05 10.96 (-3.74) ( 2.10) ( 2.79) Offer price revision 0.01 0.71 0.78 ( 1.23) ( 5.09) ( 9.65) Pre-IPO Mkt Avg Ret -0.02 23.36 17.11 (-0.04) ( 3.10) ( 2.15) Pre-IPO Mkt Std Ret -0.03 3.70 8.60 (-0.06) ( 0.73) ( 1.58) Log(1+Age) 0.23 -0.46 -3.75 ( 2.44) (-0.56) (-2.42) Year dummies Yes Yes No No Underpricing instrument 0.00 ( 0.54) Coverage instrument 9.76 ( 3.23) Pseudo or Adjusted R2 0.2366 0.5162 0.1728 0.4455
40
Table VII
Sub-period Results Descriptives statistics and two-stage regression coefficients for each of three subperiods, 1993-1994, 1995-1997, and 1998-2000. Panel A reports average underpricing, the percentage of issues in which the lead underwriter has an all-star analyst, and the percentage of issues for which the analyst from the lead underwriter provides a recommendation as of the one-year anniversary of the IPO. Panel B reports coefficient estimates with t-statistics in parentheses below for selected independent variables from two-stage regression models identical to those estimated in Table VI.
1993-1994
1995-1997
1998-2000
Full Sample
Panel A: Descriptive Statistics
Average underpricing 11.4% 17.7% 73.8% 27.5% % with All-star analyst 16.9% 18.2% 39.9% 22.4% % with coverage from lead underwriter
66.9% 86.3% 87.8% 79.9%
Number of IPOs 354 473 223 1050 Panel B: Coefficients from Two-Stage Regressions
All- star analyst (1st stage) 5.92 (2.37)
0.43 (2.08)
21.13 (1.84)
9.01 (2.49)
All-star analyst (2nd stage) 7.99 (1.53)
2.72 (0.79)
15.14 (1.45)
13.92 (3.64)
Coverage instrument 11.86 (2.33)
5.80 (2.05)
5.16 (0.62)
9.76 (3.23)
41
Table VIII
Switching Propensity Tabulation of IPOs by underpricing quintile and presence of a recommendation by the lead underwriter as of the one-year anniversary of the IPO. The table also shows the percentage of firms in each cell that switch underwriters for the SEO. The sample includes 1,050 initial public offerings (IPOs) completed between 1993 and 2000 for which a subsequent seasoned equity offering (SEO) is made between 1993 and 2001. p-values are reported for the significance of a test of equal switching rates across cells.
No Lead Recommendation Lead Recommendation
Underpricing Quintile
Count of Issuers
% of Issuers Switching
Underwriters
Count of Issuers
% of Issuers Switching
Underwriters
p-value
Low 53 73.58% 157 36.94% 0.0000 Q2 61 59.02% 151 39.07% 0.0081 Q3 36 72.22% 172 25.58% 0.0000 Q4 31 61.29% 179 20.11% 0.0000 High 30 40.00% 180 12.78% 0.0002 p-value 0.0251 0.0000
42
Table IX
Probability of Switching Lead Underwriters Results of a logit model predicting whether an issuer switches lead underwriters from IPO to the first SEO. The table reports the estimated coefficient and t-statistic for the test of a zero coefficient, as well as the predicted magnitude of impact on the probability of switching. Each magnitude is calculated by comparing the predicted change in probability of switching from perturbing the variable of interest while holding all other values at their sample means. For IPO or SEO Lead All-star, the perturbation is changing from zero to one. For all other variables, the perturbation is a change from the mean to the mean plus one standard deviation. Unexpected coverage is the residual (actual coverage dummy minus predicted probability of coverage) from the second-stage coverage model in Table VI, where coverage is defined as having an analyst recommendation at the one-year anniversary of the IPO. Standard errors in this regression correct for first-stage estimation error using the method in Murphy and Topel (1985). The sample includes 1,050 initial public offerings (IPOs) completed between 1993 and 2000 for which a subsequent seasoned equity offering (SEO) is made between 1993 and 2001.
Coefficient t-stat Magnitude Coefficient t-stat Magnitude Constant -0.4224 -0.11 1.6265 0.36 Log(Proceeds) -0.1285 -0.81 -0.0232 -0.1937 -0.90 -0.0345 Offer Price Revision -0.0158 -3.17 -0.0703 -0.0158 -2.45 -0.0701 Share Turnover 0.0078 0.84 0.0171 0.0076 0.39 0.0167 Spread 0.2667 1.41 0.0438 0.1510 0.73 0.0244 IPO Lead All-Star -0.0693 -0.29 -0.0147 -0.0875 -0.24 -0.0185 SEO Lead All-Star 0.2504 1.08 0.0550 0.2878 1.16 0.0632 IPO Underwriter rank -0.6446 -7.43 -0.1974 -0.6945 -6.15 -0.2060 SEO Underwriter rank 0.5214 5.87 0.1873 0.5490 5.47 0.1975 Days from IPO to SEO 0.0020 9.53 0.2029 0.0019 8.96 0.2010 Log(1+Age) -0.1231 -1.53 -0.0256 -0.0941 -1.12 -0.0196 Underpricing -0.0037 -1.22 -0.0392 -0.0033 -1.03 -0.0351 Unexpected Coverage -1.0154 -4.75 Pseudo R2 0.2644 0.2816
43
Underpricing Quintile Coverage0
10
20
30
40Panel A: Average Underpricing
No Yes
Underpricing Quintile Coverage60
65
70
75
80
85
90
95Panel B: % with Coverage
1 2 3 4 5
Underpricing Quintile Coverage0
10
20
30
40
50Panel C: % with All−Stars
1 2 3 4 5 No Yes
Underpricing Quintile Coverage0
10
20
30
40
50
60
70Panel D: % Switching Underwriter
1 2 3 4 5 No Yes
Figure 1. The sample is partitioned into quintiles based on underpricing, and into two groups on the basis of whether or not the company receives analyst coverage. The figure then depicts average underpricing, the percentage of companies with analyst coverage, the percentage of companies in which the lead underwriter has an all-star analyst covering the company’s industry, and the percentage of companies switching underwriters between their IPO and their SEO within each group. The full sample includes 1,050 initial public offerings (IPOs) completed between 1993 and 2000 for which a subsequent seasoned equity offering (SEO) is made between 1993 and 2001.
44
Endnotes * The authors gratefully acknowledge the contribution of Thomson Financial for providing earnings per share
forecast data, available through the Institutional Brokers Estimate System. This data has been provided as a part of a
broad academic program to encourage earnings expectations research. We thank Raj Aggarwal, Mike Cooper,
Diane Denis, Rob Hansen, Greg Kadlec, Laurie Krigman, Alexander Ljungqvist, Tim Loughran, Michelle Lowry,
John McConnell, Raghu Rau, Jay Ritter, Per Stromberg, an anonymous referee, and seminar participants at
Concordia University, Michigan State University, the University of Pittsburgh, and the 2nd Conference on
Entrepreneurship, Venture Capital, and IPOs, for helpful comments. We also thank Laura Field and Jay Ritter for
providing data, and Matt Barcaskey, Valeriy Sibilkov, and Mira Straska for research assistance.
1 For example, Das, Guo, and Zhang (2002) report the following quote from Todd Wagner, former CEO of
Broadcast.com, on the company’s decision to hire Morgan Stanley as the lead underwriter in its 1998 IPO. “Our
rationale was, if we went with Morgan Stanley, we’d get Mary Meeker (star analyst), and we’d get a lot of
attention.”
2 Whether such research is indeed valuable is open to debate. For recent evidence on the information content of
analyst research reports, see Mikhail, Asquith, and Au (2002) and Jegadeesh, Kim, Krische, and Lee (2002).
3 Hakenes and Nevries (2000) make a similar argument for IPO underpricing, while Grullon, Kanatas, and Weston
(2003) show that firm visibility (as measured by product market advertising) increases liquidity.
4 At the time of an IPO, insiders typically commit to a lock-up provision that restricts them from selling their shares
for 180 days following the IPO without the explicit written permission of the lead underwriter.
5 The proposed Rule 2712 can be found at www.nasdr.com/pdf-text/0255ntm.pdf.
6 In one well-publicized case, CSFB is alleged to have allocated an additional 15,450 shares of VA Linux Systems’
IPO to Ascent Capital based on Ascent’s recent and expected future trading activity. Based on the record 698%
increase in VA Linux’s shares on the first day of trading, Ascent’s total allocation of shares produced paper profits
of $3.8 million. That same day, Ascent traded large blocks of shares in several stocks through CSFB at
45
commissions far higher than normal. For example, Ascent is alleged to have paid $2.70 per share to trade 50,000
shares of Citgroup, a trade that would normally be done for fees of a few cents per share. See “At CSFB, Lush
Profits from IPOs Found Their Way Back to Firm,” Wall Street Journal, November 30, 2001.
7 Merrill Lynch is not covered in the I/B/E/S database prior to 1998. For offers in 1996 and 1997, we are able to
identify whether Merrill Lynch provides analyst coverage by hand collecting data from Investext. However, these
data are not available prior to 1996. In order to avoid mislabeling some Merrill Lynch-led IPOs as having no analyst
coverage, we exclude all Merrill offers for which the one-year anniversary of the IPO occurs prior to 1996. Our
results are not sensitive to this choice. In addition, we verify that other major underwriters are covered by I/B/E/S
for our entire sample period.
8 We also measure whether the lead underwriter provides an earnings forecast during the year following the IPO,
and whether the lead underwriter provides either a recommendation or a forecast. Banks that have stopped coverage
one year post-IPO, but covered the firm before or after the one-year mark are counted as not receiving coverage. In
the former case, we argue that the coverage is not ongoing, while in the latter case, we argue that the coverage is not
timely.
9 We recategorize the Institutional Investor industry definitions. For example, they consider Managed Care and
Health Care Facilities separately, while we aggregate these into a single Health Care industry, SIC 80xx.
10 We thank Jay Ritter for making these and other data available on his website (http://bear.cba.ufl.edu/ritter/). If
there are multiple lead managers we use the average reputation measure. The volume and underpricing series used
are those including all IPOs, including penny stocks.
11 The low figure in 2000 is due in part to our requirement that the firm also complete an SEO by December 2001.
In Section III.G., we provide evidence that our results are robust to the exclusion of IPOs completed in 1999 and
2000.
46
12 Typically, this means that a bank listed on SDC is matched to an I/B/E/S bank. However, it also includes a few
cases in which the SDC bank is known not to make recommendations (e.g., Allen & Co).
13 To help interpret the meaning of this ranking, BB&T and Legg Mason are rated 7, while Bear Stearns and UBS
Warburg are rated 8.
14 Consistent with Bradley, Jordan, and Ritter (2003), this is more common in the earlier years of our sample period.
Of the 117 IPOs in which there is no coverage by the lead underwriter, but there is coverage by non-leads, 67 are
completed between 1993 and 1995. Only 14 are completed in 1999 or 2000. Similarly, Bradley, Jordan, and Ritter
(2003) report that 209 of the 496 IPOs completed in 1996 did not have immediate initiation of analyst coverage,
while this was true for only 12 of the 273 IPOs completed in 2000.
15 As pointed out by Habib and Ljungqvist (1998), underpricing is mechanically related to offer size. Thus, the
interpretation of this variable as a proxy for uncertainty is problematic. We include it in order to facilitate
comparison of our findings to those of prior studies and to control for possible economies of scale in underwriting.
In unreported regressions, we also measure issue size as the log of expected proceeds, where expected proceeds is
equal to the midpoint of the original filing price range times the number of shares offered. Our results are virtually
identical using this alternate size measure.
16 We assume that the offer price revision is exogenous. Ljungqvist and Wilhelm (2002) and Benveniste et al.
(2003) model the revision as an endogenous variable.
17 This first result is slightly biased since our measure of average underpricing across all IPOs includes the specific
IPO being analyzed. However, this bias will be quite small given the large number of IPOs per month over our
sample period.
18 We correct for estimation error induced by the generated regressor using Equation (34) in Murphy and Topel
(1985).
19 In untabulated results, we also include a variable measuring the annualized stock return between the IPO and the
SEO. This variable is statistically insignificant and does not affect the significance of the other independent
variables.
47
20 Recall from Table I that we are able to link the lead underwriting bank from SDC with an analyst firm from
I/B/E/S in only 94% of the cases in 1993 and 85% of the cases in 1994. This percentage jumps to 99% in 1996. Of
the 42 cases in which the lead underwriter has an all-star analyst, but for which we have no record of an analyst
recommendation at the one-year anniversary of the IPO, fourteen occur in 1993. This is consistent with some of
these cases being due to data errors induced by incomplete I/B/E/S coverage in 1993 and 1994.
21 It is possible that the cases in which we observe earnings estimates, but no recommendations are I/B/E/S data
errors. There are two reasons why we doubt that such errors are pervasive. First, the cases are not restricted to the
early part of the sample period when I/B/E/S coverage was less complete. Ten of the 98 cases are from IPOs
completed in 1999 or 2000. Second, we hand-checked a number of these cases with other data sources such as
Investext, and did not uncover systematic problems with the I/B/E/S data. Of course, we can’t completely rule out
the possibility of some data errors. However, we note that, in order for such data errors to be driving the positive
association between coverage and underpricing, it would have to be the case that those cases with errors were
systematically less underpriced than the others. We can think of no reason why this should be true.
22 See, for example, “Some Analysts Leave Industry in Search of ‘New Adventure,’ Wall Street Journal Online,
February 28, 2003 and “Miffed, Four CSFB Analysts Depart: Angered by Skimpy Bonus Payments, Healthcare
Quartet Signs on at B of A,” Investment Dealers Digest, March 3, 2003.